Commit 61a5960c authored by Yuxin Wu's avatar Yuxin Wu

docs about distributed data (fix #202)

parent cddb713f
...@@ -198,11 +198,31 @@ The above DataFlow can run at a speed of 5~10 batches per second, if you have go ...@@ -198,11 +198,31 @@ The above DataFlow can run at a speed of 5~10 batches per second, if you have go
As a reference, tensorpack can train ResNet-18 (a shallow ResNet) at 4.5 batches (of 256 samples) per second on 4 old TitanX. As a reference, tensorpack can train ResNet-18 (a shallow ResNet) at 4.5 batches (of 256 samples) per second on 4 old TitanX.
So DataFlow won't be a serious bottleneck if configured properly. So DataFlow won't be a serious bottleneck if configured properly.
## Larger Datasets? ## More Efficient DataFlow
For larger datasets (and smaller networks) you could be seriously bounded by CPU or disk speed of a single machine. To work with larger datasets (or smaller networks, or more GPUS) you could be seriously bounded by CPU or disk speed of a single machine.
Then it's best to run DataFlow distributely and collect them on the Then it's best to run DataFlow distributely and collect them on the
training machine. Currently there is only little support for this feature. training machine. E.g.:
```python
# Data Machine #1, process 1-20:
df = MyLargeData()
send_dataflow_zmq(df, 'tcp://1.2.3.4:8877')
```
```python
# Data Machine #2, process 1-20:
df = MyLargeData()
send_dataflow_zmq(df, 'tcp://1.2.3.4:8877')
```
```python
# Training Machine, process 1-10:
df = MyLargeData()
send_dataflow_zmq(df, 'ipc:///tmp/ipc-socket')
```
```python
# Training Machine, training process
df = RemoteDataZMQ('ipc:///tmp/ipc-socket', 'tcp://0.0.0.0:8877')
TestDataSpeed(df).start_test()
```
[1]: #ref [1]: #ref
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment